The update introduced multimodal embeddings that encode visual, speech, and video cues beyond conventional text transcriptions, facilitating richer search and retrieval experiences from media assets stored in Amazon S3. Marengo 3.0 generates compact 512-dimensional vectors, improving accuracy and enabling users to navigate directly to relevant video segments.
This advancement expands use cases in fields like sports analytics and education, where precise segment retrieval by concept is important. The configurable segmentation matches content structures, and integration requires no user-managed infrastructure, streamlining media search within Bedrock environments.
Source & further reading
Amazon Bedrock Managed Knowledge Base now supports multimodal embeddings for video, audio, and image content with TwelveLabs Marengo 3.0 ↗Source published 2026-09-11. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.